Motion capture involves recording the position and global orientation of joint sensors of a real object, in most cases a real person performing some human activities. This information is usually recorded at uniformly spaced instances of time, or as it is often called frame-by-frame. Then the recorded motion data sets are processed and mapped into a skeleton hierarchy of a virtual, computer simulated human figure to control the motion of the virtual human in the computer simulation. In the first part of the paper we review several new techniques developed to facilitate the manipulation, noise reduction, storage and reuse of captured data, which have a potential to reduce the overall cost of motion simulation and improve its realism. In the second part of the paper we consider the real life problem of reducing a worker’s risk from being hit by underground mining machinery in a confined space. We formulate a set of requirements to motion editing for this particular task and analyze the limitation of existing techniques

National Institute for Occupational Safety and Health